NDT.net • June 2006 • Vol. 11 No.6

Nondestructive Testing of Meat Using Planar ElectroMagnetic Sensors

S. C. Mukhopadhyay and C. P. Gooneratne,
Institute of Information Sciences and Technology
College of Sciences
Massey University (Turitea)
Palmerston North, New Zealand
Email: S.C.Mukhopadhyay@massey.ac.nz


Abstract

The paper has described the possibility of testing meat in a non-destructive and non-invasive way using planar electromagnetic sensors. The planar electromagnetic sensor of interdigital type is quite sensitive to the fat content of meat samples. The feasibility of developing a low cost instrumentation system for estimation of fat content in a nondestructive way has been explored.
Keywords: Nondestructive testing (NDT), non-invasive inspection, planar electromagnetic sensors, interdigital type, impedance, meat inspection, fat content.

1 Introduction

The purpose of this research activity is to explore the possibility of using planar type electromagnetic sensors for testing quality of meat especially the fat content in meat samples in a nondestructive way. Earlier research using planar magnetic sensors has demonstrated the detection of defects in the printed circuit board and the estimation of near-surface material properties of conducting and magnetic materials. This sensing technique has the potential to be used for the quality inspection of milk products such as milk, butter, cheese, curds, yogurts etc. This sensing technique is based on a planar electromagnetic sensor operating with radiofrequency excitation and the computer-aided computation being fast allows the on-line monitoring of quality. The on-line sensing system suitable for the dairy industry needs to possess several key qualities including cost effectiveness, high reliability both in terms of measurement accuracy and speed of measurement. Also the sensor technology has to be able to perform volumetric penetrative measurement in order to measure properties through out the product and this could be achieved by employing a planar type electromagnetic sensing technique. In this research planar electromagnetic sensors have been investigated for the nondestructive and non-invasive determination of fat content of meat samples especially pork belly cuts and some results of initial investigations are reported in this paper. The development of a planar-electromagnetic-sensors-based inspection system can be an alternative to chemically destructive methods.

2 Planar Electromagnetic Sensors

The research and development works using planar electromagnetic sensors have started quite some time back. The works employing planar meander type sensors started quite a few years back with the target of developing a complete inspection system of the printed circuit board (pcb) for pentium pr°Cessors [1]. The exciting coil is of meander configuration as shown in figure 1 whereas the sensing coil is either of a mesh type or a figure-of-eight type configuration. The pcb of pentium pr°Cessor having many long conductors, the meander configured exciting coil is a suitable choice. The planar meander type sensor has also been used for the inspection of material defects such as the existence of inner layer cracks and for the estimation of fatigue of metal products [2, 3, 4, 5]. A crack with alignment in parallel with the exciting meander coil is difficult to be detected by the meander configuration. The alternative is to employ mesh type sensors as shown in figure 2 [6, 7, 8, 9]. The response of both the meander and mesh type planar electromagnetic sensors to dielectric materials is moderate at relatively low frequency of excitation. In order to increase the sensitivity of the sensor system another type of sensor, the interdigital one as shown in figure 3 has been fabricated and developed. Planar interdigital type sensors have been used for many different applications; planar interdigitated array electrodes have been used for the determination of components contained in aqueous solutions [10], determination of moisture, fiber and titanium dioxide in paper pulp [11, 12, 13], for the complex permittivity characterization of materials [14] and the detection of small amount of magnetic beads over the surface of microsensors [15].

Both the meander and mesh type sensors consist of two coils: one used for excitation to generate electromagnetic field and another for sensing. The high frequency alternating supply is provided to the exciting coil and the voltage across the sensing coil is measured. The transfer impedance (the ratio of the sensing voltage to the exciting current) is used as the characterization parameter for the meander and mesh type sensors. In case of interdigital sensor the exciting voltage and the displacement current through the sensor is measured. The impedance (the ratio of the applied voltage to the current) is the characterization parameter used for the interdigital sensor. To have a knowledge of the impedance characteristics the impedances have been measured experimentally. The experimental set-up is shown in figure 4.


Figure 1: Fabricated meander type planar electromagnetic sensor

Figure 2: Fabricated mesh type planar electromagnetic sensor

Figure 3: Fabricated interdigital type planar electromagnetic sensor


Figure 4: The experimental set-up for sensor characterization

Different sizes of sensors of each meander, mesh and interdigital types have been experimented using the experimental set-up. The sensor has been supplied from a controlled frequency supply, the voltage across and the current through the sensor are recorded. The frequency of excitation has been varied between 100 kHz to 80 MHz. The impedance characteristics of the sensors are shown in figures 5, 6 and 7 respectively. It is seen that the transfer impedance for both meander and mesh type increases with the increase in frequency whereas the impedance of interdigital type sensor decreases with frequency. With the same effective area the response of mesh type sensors is better than meander type. Basically meander and mesh types sensors are inductive types whereas the interdigital one is a capacitive type. It is also seen that both meander and mesh type sensors respond well at high frequency whereas the response of the interdigital sensor is very good at low frequency and doesn’t respond well at high frequency. The choice of operating frequency must be carefully selected.


Figure 5: Transfer impedance characteristics of meander type sensor

Figure 6: Transfer impedance characteristics of mesh type sensor

Figure 7: Transfer impedance characteristics of interdigital type sensor

3 Initial Investigation with Planar Interdigital Sensor

Since the response of the planar meander and mesh type sensors with dielectric material such as pork meat is not good at all, planar interdigital sensor has been used for the experiment. A planar inter-digital sensor (40mm׳40mm) as shown in figure 3 has been used to conduct experiment with different types of pork using the experimental set-up as shown in figure 4. There were mainly four main varieties
  • Fat
  • Mixed
  • Muscle
  • Skin.
The sensor was tested on 4 pieces of fat, 2 pieces of mixed meat, 4 pieces of muscle and 1 piece of skin. One piece of each type is shown in figure 8. All the pieces were of different shape and size. The experiment involved analyzing the sensor results, to see if the different amount of fat content affect the impedance of the sensor.

Each piece was tested separately, and the sensor was placed as close as possible to the meat for all experiments except experiment 1. In experiment 1 glad-wrap was placed on top of the meat, and the sensor lightly touched the top of the meat. During the last experiment (exp 4) the skin was put on top of a piece of fat, mixed and muscle. With the exception of experiment 1, the sensor had no direct contact with the meat, during all other experiments.


Figure 8: Pork samples for test

The inter-digital sensor was tested with a series resistor across which the voltage is measured to get the current signal. The operating frequency was set at 500 KHz. The impedance was then calculated as shown below. Voltage across the exciting coil = V1
Current (i) across sensing coil, i = V2/R (R is the series resistor, V2 is the voltage across it)
Impedance = V1/i = R*(V1/V2).
Temperature noted as 16.1 °C.

A few experimental results are shown here. Figure 9 shows a typical signals obtained. The signal 1 corresponds to the applied input voltage and signal 2 corresponds to voltage obtained across the series resistor.


Figure 9: Typical signals from experiment

Figure 10: Impedance of the sensor obtained from Experiment 2

The impedance is calculated for all experiments and the variation of impedance of the sensor with different conditions are shown in figure 10 corresponding to experiment 2.

An initial measurement is obtained for the sensor with no materials under it, before each of the four experiments. In figure 10 this measurement is shown as “Air”. Looking closely at figure 10 it is seen that the impedance of the sensor is different for different samples. This shows that the sensor can detect different types of pork. This is due to different effective permittivity of different pork meats.

The sensor used in the experiment is not the optimum one required for the best result. But observing the results it can be concluded that the planar interdigital sensor is sensitive to pork meat and has the potential to be used for the estimation of fat content in pork meat. More experiments are required to determine the actual nature of the influence of fat on sensor characteristics.

4 2nd Set of Experiment

Based on the above observation it has been realized that the planar electromagnetic sensor of interdigital configuration has the potential to determine the fat content of pork belly cuts. In the initial investigation the size of the sensor was not sufficient to cover the complete piece of meat. In practice the bellies are cut into particular sizes and the possibility of testing them by doing experiment only one are preferred. To explore that possibilities three new sensors of different pitch sizes have been fabricated keeping the covered area almost equal, the details are given in Table 1. They were tested on six pork belly pieces, A1, B1, C1, A2, B2, and C2. the bellies were around 20 mm – 30 mm deep with skin on and were placed on top of each sensor to simulate the factory situation. The pieces of pork bellies were placed on the sensor according to four different orientations as explained below with respect to figure 11.

Table 1: Details of sensors
Sensor Periodicity(mm) Fingerlength( mm)
1 15 120
2 20 120
3 30 120


Figure 11: Representation of pork belly cut

Orientation 1 = [skin side up, label at front] – as shown in figure 11.
Orientation 2 = [skin side up, label at back] – Rotate 180°.
Orientation 3 = [skin side down, label at front but underneath] – Flip over.
Orientation 4 = [skin side down, label at back underneath] - Rotate 180°.

The sensors have one sided access to the meat under test (MUT). Electric field lines pass through the MUT, and the capacitance between the two electrodes, depend on the material dielectric properties as well as on the electrode and material geometry.

The sensors were driven by a 10 V Sine wave. The measurements were made at frequencies in the range from 5 KHz to 1 MHz. The pork belly pieces had skin on top and muscle at the bottom, where the personal view of top and bottom is contradictory. The sensors were rested on a table with an insulating mat underneath, with the electrodes facing up. Glad wrap was placed on top of the sensor to prevent direct contact with the pork. This is done to keep up the high standards of hygiene required when testing meat. Each of the six pieces of pork was tested for all four orientations at the same frequency range. The driving signal for the sensors were provided by the Agilent 33120A waveform generator and the Agilent 54622D mixed signal oscilloscope, analyzed the input voltage, output current and the phase. All efforts were made to make sure the pork pieces were all tested at similar temperatures, varying between 16-18°C.

A few results are reported here. Since different pork belly samples may have different effective permittivity, the imaginary component of impedance will be mainly affected. Figure 12 shows the variation of reactive part of impedance of sensors #1 as a function of frequency for all six samples of pork belly for the orientation#1. It is seen that there is a distinct difference in the magnitude of impedance in the frequency range 5 kHz to 40 kHz. With an increase in frequency the difference in magnitude between samples decreases.


Figure 12: Variation of reactive impedance of sensor#1 with frequency for six different pork belly (orientation#1)

Figures 13 and 14 show the variation in reactive impedance for sensors #2 and #3, respectively, as a function of frequency for the same orientation#1. It can be seen that the results obtained for sensors#2 and #3 are quite different from that obtained for sensor#1. Figures 15, 16 and 17 show the reactive impedance as a function of frequency for all three sensors for orientation #3. Even though the results are quite distinct for all six samples, the results obtained corresponding to orientation#1 are quite different to orientation #3. This is due to the fact that the response of the sensors depend on its penetration depth. If the thickness of pork belly is larger than the penetration depth of field lines, the sensors may not be able to respond to fat if the fat lies on the top surface of the meat piece. The impedance of the sensors provides an average indication of the fat to the depth of penetration and the part of the meat within the electric field.


Figure 13: Variation of reactive impedance of sensor#2 with frequency for six different pork belly (orientation#1)

Figure 14: Variation of reactive impedance of sensor#3 with frequency for six different pork belly (orientation#1)

Figure 15: Variation of reactive impedance of sensor#1 with frequency for six different pork belly (orientation#3)

Figure 16: Variation of reactive impedance of sensor#2 with frequency for six different pork belly (orientation#3)

In practice, measurement for one frequency or at only a few frequencies are required. Figure 18 shows the reactive impedance values of different samples at an operating frequency of 5 kHz corresponding to orientation #1. It is seen that the impedance values are quite distinct from each other, the sample A1 has got the highest impedance and the sample A2 has the lowest impedance. The measurement at 5 kHz provides the opportunity of development of a low cost instrumentation and measurement system based on a Cygnal C8051F020 microcontroller.


Figure 17: Variation of reactive impedance of sensor#3 with frequency for six different pork belly (orientation#3)

Figure 18: Variation of reactive impedance of sensors#1 for different pork belly cuts at operating frequency 5 kHz for orientation #1

The key question is whether the reactive impedance values shown in figure 18 have a relationship with the fat content of the pork belly. To compare the performance of the sensors 3 test samples were analysed by Soxhlet extraction of an homogenized sample (including the skin) using petroleum ether (Bp 40 – 60°C) and the results are shown in Table 2. The maximum fat content is for the sample A1 which is obtained from the results corresponding to sensor #1. From figure 18 it is seen that the average fat content of B2 is slightly higher than A2. From the results obtained from the sensors it can be concluded that the planar interdigital sensors has a good potential for the on-line determination of fat content of pork meat.

Table 2: Fat content from chemical analysis
Chemi cal Test Sample #A1 Sample #A2 Sample #B2

30.09 19.96 17.36

The mathematical part for the estimation of fat content is described here. The magnitude and phase of the impedance of the sensor without meat (air) and with meat under test (MUT) are measured. The reactive impedances are calculated as follows
Xair = Zair *sin(fair); Zair and fair are the impedance magnitude and phase of the sensors without meat.
XMUT = ZMUT *sin(fMUT); ZMUT and fMUT are the impedance magnitude and phase of the sensors with pork belly under test.
The effective permittivity of the sample is calculated as air
The inverse of the effective permittivity is taken as the parameter of index, ., and is used for the analysis to determine the fat and protein content.
For the calculation of fat the following equation is used
Fatcal = 48.1 * (k - 0.15) + 18.1
and for the calculation of protein the following equation is used
Proteincal = 16.5 - 16.1 * (k - 0.15).
The parameters of the above equations are obtained from two test sample which corresponds to calibration of the sensor. Based on the above equation the fat and protein content of the samples are obtained and is shown in Table 3.
It is seen from Table 3, that the predicted results are very close to the experimental one. In practice it is impossible to get exactly the same results using a planar sensor to that obtained from chemical analysis. The planar sensor provides an average result of a large sample of pork belly whereas the result from chemical analysis are based on 5 gm sample though it is homogenized before the sample for analysis is taken.

Table 3: Estimation of fat and protein content
Sample Index parameter (k) Calculated Fat content Calculated protein content
A1 0.3987 30.06 12.49
A2 0.1633 18.74 16.28
B1 0.2353 22.20 15.12
B2 0.1644 18.80 16.26
C1 0.3053 25.57 13.99
C2 0.2005 20.53 15.68

5 3rd Set of Experiment

A further experiment has been carried using the three sensors on thirteen pork belly samples. The new samples are smaller and more shallow pork belly samples with skin removed so they would correspond more closely to the electric field penetration. Like the earlier experiment the readings were taken for four different orientations. Figures 19 to 21 show the variation of reactive part of the impedance of three different sensors as a function of the frequency for orientation#1. The results are shown upto a maximum operating frequency 30 kHz. With the higher frequency the difference in readings between samples are not noticeable.


Figure 19: Variation of reactive impedance of sensor#1 with frequency for thirteen different pork belly (orientation#1)

Figure 20: Variation of reactive impedance of sensor#2 with frequency for thirteen different pork belly (orientation#1)

Figures 22 to 24 show the variation of reactive part of the impedance of three different sensors as a function of frequency for orientation#3. The results are shown upto a maximum of operating frequency of 30 kHz. With the higher value of operating frequency the difference in readings between samples are not noticeable. The results are slightly different than those obtained for orientation#1.


Figure 21: Variation of reactive impedance of sensor#3 with frequency for thirteen different pork belly (orientation#1)

Figure 22: Variation of reactive impedance of sensor#1 with frequency for thirteen different pork belly (orientation#3)

Figure 23: Variation of reactive impedance of sensor#2 with frequency for thirteen different pork belly (orientation#3)

Figure 24: Variation of reactive impedance of sensor#3 with frequency for thirteen different pork belly (orientation#3)

Figures 25 and 26 show the reactive part of the impedances of sensor#1 corresponding to orientation#1 for thirteen different samples at operating frequencies 5 kHz and 10 kHz, respectively. It can be seen from figures 25 and 26, that the variation between samples for these two frequencies are exactly same. This shows that the results obtained from frequencies as low as 5 kHz provide quite distinctly different results for different samples.


Figure 25: Variation of reactive impedance of sensors#1 for different pork belly cuts at operating frequency 5 kHz

Figure 26: Variation of reactive impedance of sensors#1 for different pork belly cuts at operating frequency 10 kHz

6 Cygnal 8051 Microcontroller-based instrumentation system

For the analysis it is necessary to collect voltage and current signals, so an efficient data acquisition system is very important. The analog data is captured using an analog-to-digital converter. A SiLab microcontroller C8051F020 was considered initially as shown in figure 27. The SiLab C8051F020 has two ADCs operating at 100 kHz and 500 kHz respectively. So an operating frequency of up to 50 kHz can very well be used using this system. The system is under development and is expected to provide good results.


Figure 27: SiLab microcontroller C8051F020 based data acquisition system

7 Conclusions

This paper has described the interaction of planar electromagnetic sensors with samples from pork belly cuts. The sensors of planar interdigital configuration of different pitches have been fabricated and used for experiments. The sensors show quite distinctive results with different pork belly samples. The response of the sensors indicate a good possibility of using this type sensor for the prediction of fat content in pork belly in an non-invasive way. The good response of the sensors at low frequencies suggests that it should be possible to develop a low cost sensing system based on a Cygnal microcontroller C8051F020.

8 References

  1. Yamada, S., Fujiki, H., Iwahara, M., Mukhopadhyay, S.C., and Dawson, F.P., "Investigation of Printed Wiring Board Testing By Using Planar Coil Type ECT Probe", IEEE Transactions on Magnetics, Vol. 33, No. 5, Sept, 1997, pp 3376-3378.
  2. Goldfine, N.J., “Magnetometers for improved material characterization in aerospace application”, Material Evaluation, March 1993, pp. 396-405.
  3. Goldfine, N.J., Clark, D., and Lovett, T., ‘Material characterization using model based meandering winding eddy current testing (MW-ET),’ EPRI Topical workshop: Electromagnetic NDE applications in the Electric Power Industry, Aug 21-23, 1995, Charlotte, NC.
  4. Goldfine, N.J., ‘Conformable, meandering winding magnetometer (MWM) for flaw and material characterization in ferrous and nonferrous metals,’ ASME Pressure Vessels and Piping Conference, Pr°Ceeding on International Advancement in PVP Technology, July 27- 31, 1997, Orlando, FL, USA.
  5. Mukhopadhyay, S.C., “Quality inspection of electroplated materials using planar type micro-magnetic sensors with post pr°Cessing from neural network model”, IEE Pr°Ceedings – Science, Measurement and Technology, Vol. 149, No. 4, pp. 165-171, July 2002.
  6. Mukhopadhyay, S.C., Yamada, S., and Iwahara, M., “Experimental Determination of Optimum Coil Pitch for a Planar Mesh Type Micro-magnetic Sensor”, IEEE Transactions on Magnetics, Vol. 38, No. 5, Sep. 2002, pp 3380-3382.
  7. Mukhopadhyay, S.C., Yamada, S., and Iwahara, M., “Optimum coil pitch selection for planar mesh type micro-magnetic sensor for the estimation of near-surface material properties”, JSAEM studies in Applied Electromagnetics and Mechanics, Vol. 14, pp 1-9, 2003.
  8. Mukhopadhyay, S.C., “A Novel Planar Mesh TypeMicro-electromagnetic Sensor: Part I - Model Formulation”, IEEE Sensors Journal, Vol. 4, No. 3, pp. 301-307, June 2004. [9] Mukhopadhyay, S.C., “A Novel Planar Mesh Type Micro-electromagnetic Sensor: Part II – Estimation of System Properties”, IEEE Sensors Journal, Vol. 4, No. 3, pp. 308-312, June 2004.
  9. K. Toda, Y. Komatsu, S. Oguni, S. Hashiguchi and I. Sanemasa, “A Planar Gas Sensor Combined with Interdigitated Array Electrodes”, Analytical Sciences, The Japan S°Ciety for Analytical Chemistry, January, 1999, Vol. 15, pp. 87-89.
  10. B.H.Timmer, W. Sparreboom, W. Olthuis, P. Bergveld and A. van den Berg, “Planar Interdigitated Conductivity Sensors for Low Electrolytic Concentrations”,
  11. K. Sundara-Rajan, L. Byrd, and A. V. Mamishev, "Measuring Moisture, Fiber, and Titanium Dioxide in Pulp With Impedance Spectroscopy," TAPPI Journal, vol. 4, no. 2, pp. 23-27, Feb. 2005.
  12. K. Sundara-Rajan, L. Byrd, and A. V. Mamishev, "Moisture Content Estimation in Paper Pulp Using Fringing Field Impedance Spectroscopy," Sensors Journal, IEEE, vol. 4, no. 3, pp. 378-383, 2004.
  13. E. Fratticcioli, M. Dionigi and R. Sorrentino, “A Planar Resonant Sensor for the Complex Permittivity Characterization of Materials”, IEEE MIT-S Digest, 2002, paper no. WE1B-3, pp. 647-649.
  14. S. Baglio, S. Castorina and N. Savalli, “Integrated Inductive Sensors for the Detection of Magnetic Microparticles”, IEEE Sensors Journal, Vol. 5, No. 3, June 2005, pp. 372-384.

© NDT.net |Top|